Benji Rossman — Associate professor at Wits University, Head Of The RAIL Lab, Robotics, Autonomous Intelligence and Learning

Monday night, what's involved as we do each and every Monday? Good to have you along with us. Stick's name, Mr. Roboto. I thought it was appropriate, considering my first guest in the studio tonight. And he's probably going to have to help me do all of those wonderful things, like explain exactly what he does do. But it's Benji Rossman. Hi, Benji. Hey. Good to have you. OK, so you're at VIT at the moment. Yes. I'm just going to-- oh, excuse me while I beat myself to do the microphone. You are the head of RAIL over there. Now, initially, when I saw RAIL, I thought, OK, the guy works on railroads and railjacks. Can you sort out the transport? But it's not. What does RAIL stand for? So that's our research lab at VIT, and that's the Robotics, Autonomous Intelligence, and Learning Lab. So we're a research lab that care about everything AI, machine learning, and robots. And so you have just been made. We're allowed to announce this. And if it is, you know, official or unofficial, I'll take the blame for it. You've just been made an associate professor. Yes. So I mean, that's brilliant. At VIT, so now, what were you doing before then, if you were not an associate? Were you a lecturer? How did they work things in academia?

So I had a joint position. I was a senior lecturer at VIT, but my main position was at the CSR, which is the Council for Scientific and Industrial Research. And there was a principal researcher in robotics there as well. OK. And now you're now just going to concentrate full time, or you're still doing work with CSR. No, I'm moving full time to VIT, so. OK. Cool. All right. You're also one of the founders of the deep learning in DARBA, which is an entirely African thing, and it has done some amazing stuff. We'll talk about that. But your specialty, your lab, is robotics, machine learning, that kind of thing. Let's start off at the very beginning, though. Before we get into all of the exciting questions I want to ask you, tell me a little bit about Benji. I mean, are you a Joburg boy born in bed? Let me know a little bit because I'm always fascinated what leads people to become and to get into the careers that they've gotten into. So what is your story? Cool. So I grew up in Joburg, and I went to VITs and did my undergraduates in computer science and applied maths. And I really got into that because back at school, I was really into computer games. And I was fascinated by this idea that I didn't just have to play them.

I could build them as well. And I had this whole avenue where I could build these virtual worlds with characters and so on. And so it seemed like computer science was a natural thing to go into. And then once I'd done that, I started getting wooed over by some of the magic of artificial intelligence and machines that can learn. And I ended up deciding I wanted to study further. And I did that in the UK. I went to the University of Edinburgh and did my master's in artificial intelligence. And at that point, I was kind of hooked. So I did a PhD there as well. And I kept making my decisions based on whatever I thought was coolest at the time. And this is where I ended up. I love that. That's a very good decision-making process. But now, I actually met you because you're also on the faculty here in South Africa of Singularity U. So I met you at the summit. In fact, you were sitting in front of me. And I just barged in and introduced myself as us radio people are wanting to do. I'm glad you did. You didn't strike me as the typical academic and professorial type. And maybe it's because I've got still an old-fashioned view of what academics and professors should look like. Because you typically expect the guy in the tweed jacket with the leather patches on the elbow, maybe the smoking.

You're nothing like that. I mean, you're kind of young, dynamic, fun, trendy, cool. You're a geek, which in my book is great. I'm a big fat geek myself. And I think it's certainly bred and built some new life. And you reckon-- we were talking off-air. And you were saying sometimes the guy in the hoodie could well be a tech billionaire these days and not a gangster or a drug dealer. So tell me now, you were fascinated by this. You were fascinated. You talk about machine learning. And then we talk about artificial intelligence. Is that one and the same? I mean, a lot of people treat them that way. I like to distinguish that AI or artificial intelligence is-- we think of as a broad umbrella term for any science or engineering discipline that's about trying to build machines that act intelligently in some way. And there's a lot of different sub-areas that fall under that. But the powerhouse of it at this point is machine learning. And machine learning is just this idea of building what we call algorithms, which are methods or programs that get better with experience. So they get better over time. OK, now give me an example for somebody like myself that looks at machine learning and goes, what does that mean?

Is my fridge going to be able to-- I don't know. So give me an example of machine learning. So maybe some of the easiest or most common examples would be things like-- if you're looking at, say, vision systems that can do facial recognition. So you can look at a photo and pick up where the face is in there. Or maybe you take a photo and it says the eyes were closed in this one. So there's some insights that's going on there that is-- so it's learning over time. It's never seen those specific examples. But it's learned this idea of what maybe faces look like or what open eyes look like. And so that was done by giving it examples that it could train from. Other kinds of examples might be language translation systems. So if you go to something like Google Translate and type something in one language and it comes up in another language, those have been learning systems that have figured out how to do that. Now, what's happening a while ago? I don't know if I'm getting my Douglas Adams mixed up with reality here, but a program called Babelfish, which was also a translation-type program. But when I-- and that was a good couple of years ago. When I saw it, it was horrible. I mean, it used to be like a joke thing because it just couldn't.

Have we progressed now that it's-- I mean, Siri still doesn't-- I mean, she interrupts me at the most inopportune moment. When I want to speak to her, I can go hay Siri until I'm blue in the face, and she'll ignore me. And I'm saying she because my Siri has a female voice. So maybe it is. Maybe she's got some female personality traits, and she's going to nag me again, so I'm going to ignore him. It has improved Siri as an example, then, of machine learning. OK. So there'll be different kinds of technologies going on there, but there's definitely learning aspects, whether they're learning about language or learning about human voices. Different components of the system will be running machine learning methods. OK. Well, when we come back, we are chatting to Benjy Rossman, who is the-- what do I now call you? The Associate Professor, Head of Robotics-- how? Can we just say Benjy? You can just say Benjy. OK. We'll be back with Benjy in just a little bit. And then I want to find out about the whole-- all these conspiracy theories, the rise of the machines, et cetera, et cetera. When we come back, we'll be chatting to Benjy Rossman some more. Welcome back to What's Involved. My special guest in studio is Benjy Rossman.

OK. So now the burning question. I mean, people are asking-- oh, somebody else wants a question.

I'm going to show you this one, OK? There's two. First and foremost, when we were talking about Siri, apparently Al, who was on air before us, I spoke to his Siri in the car. This is not the first time when I've gone, hey, Siri. And people's Siri's have been activated. I don't know what it is. I thought Siri was supposed to recognize your voice. My fiance told me I nearly killed her one day because I was standing in for Tony on the breakfast show. She was going on her way into do a training. And suddenly, I was talking on the radio, and Siri was talking back to me. So maybe it's a thing. Maybe it's a thing. You made a very nice comment off of you if you were a rock star. I think that's a brilliant one. I know a couple of people that are in the music business. I must actually suggest they do that at one concert. We'll see if it works. Just shout it out from stage. Is that an alarm for 3 o'clock? We'll see. Hey, Siri. Anyway, so the other one that's Leo's just coming, he goes, evening, David, question for Benji. Can you prove that he isn't a robot? Could we be in the Matrix right now? I don't know who he's been talking to. I deny everything. I can state categorically with the utmost uncertainty that Benji is not a robot.

There you go. There we go. We can't attest to the fact that he hasn't been augmented in some way. So there could be nanites running around in his body as we know. That's what I'm going to ask him now. So the whole Matrix kind of thing, OK, yes. What people are more concerned about, and one of the questions in here is AI machine learning. Are we creating Skynet, which is one sort of fantasy and everything, or are we making that a reality? How do you feel about it? I mean, is this just the scaremongering? What is happening? I think it's funny. That's the thing we try and not talk about a huge amount, because that's often what people go to. And I think it's very much the Hollywood view of where we're going with AI. I think there's no reason to think we're creating these evil machines that are going to take over the world. And what's actually quite interesting about the way that these work and that machine learning works is that we effectively train them. So we give them examples, and that's how they get better at doing whatever they're doing. And so a lot of the work is around trying to find the right curricula for them to learn with. So you almost think of them in some way as a child, but a very specific child that's maybe in most cases doing one specific task.

But you've got to train it through various kinds of interactions. And I think the idea that just, emergently, out of nowhere, some big evil system is going to suddenly appear, that's probably not at the top of our worries list at this point. We're also a long way away from it. I would imagine we were. And I mean, a lot of people, this sort of expression, the fourth industrial revolution, a lot of people are like, it's coming, it's coming, we're all going to lose our jobs. In my opinion, it's here, number one. Number two, I don't know so much. My feeling is that with machine learning and robotics and those kind of things, it's ultimately going to give us more-- it's going to give more to us and enable us, I believe, to explore our humanity even more. What is your feeling on that? I mean, is that so far away you're like, don't even bother thinking about it? That's something I think about and I agree with. I think there's a lot of cases where-- what we talk about a lot is automating jobs. But really, we're automating aspects of jobs is where the work is really at. And practically, there's a lot of things people do in their jobs that I think they shouldn't be doing. The example that I always like is someone in HR probably went into HR because they like working with people, not because they like doing endless paperwork every day.

And so if you could minimize the amount of that stuff they had to do, you could free up their time to do what they should be focusing on. And I think this is the way to think about a lot of this kind of technology. OK, now, when we talk-- and I might be digressing a bit. But when we talk about machine learning and the internet of things, is it one and the same? Is there a convergence? Two totally different things? Where do we stand with that? So the internet of things really refers to this idea that I can effectively Wi-Fi enable any device I want, whether it's a fridge or my lights or whatever. And now, that talks more to this idea of the connectivity of your hardware. But machine learning is kind of the underlying substrate that can make that do something useful. So you've got the ability to combine these two areas. They've got different concerns, automating your gate when you drive up to it, it opens. But actually, if you want to behave more intelligently, that's where you could use machine learning to maybe learn about your behaviors and when you want certain things to happen and how different devices should interact with each other. So it's a nice platform on which to use machine learning.

All right, now, talk to me about some of the stuff that you guys are busy with, the actual application of this. Because I have this picture in my mind that the lab at Vets looks like one of these mad scientist labs with people in white coats running around, doing all sorts of crazy experiments, et cetera, et cetera. Tell me, what are you busy with? We're computer scientists. We don't like to get our hands dirty, so we don't have to wear white coats. That's why we have computers. You're not that kind of scientist. We're more like mathematicians with computers. So we work on a variety of different things. I think what I'm quite excited about is the core focus of our lab is more theoretical work. And I've always felt quite strongly about this, because there's so little of that that happens in Africa. And my view has always been that, yes, there's a lot of big problems. We should be applying a lot of these kinds of technologies to solve. But there's nothing stopping us from developing these technologies locally as well. So that's a big passion of mine. So we work on a number of theoretical questions, but also some application areas. In the theoretical space, we care a lot about the kinds of questions that are around behavior learning.

So a robot might have a behavior for making a cup of coffee, or learning to navigate around some environment, or collaborate with a person on something. It might not be a robot. It might be a software agent that's learning how to maybe trade on the stock market, or something like that. So these kinds of behaviors, and how would you go about learning these behaviors? And if you've learned to solve one problem, how do you use that knowledge to help you solve other problems? Those kinds of questions are what we look at in the Moore theoretical space. And then we also work with various questions around applications, which involve everything from questions of computer vision, which we look at in the context of, say, education. We're starting to look at questions around health care. And so there's a variety of different spaces that we're playing in at the moment. OK. I'm loving some of the comments coming in here from our listeners. One of them just in that says, Google is Skynet. There was a rolling on the floor laughing emoji. So if Google is Skynet, it's broken. It's broken, I tell you.

Here's somebody that's asking about putting a human neural network into a computer. Is such a thing even possible? I mean, that smacks of these immortality kind of things. I'll tell you, well, let's answer that question when we come back, OK? My special guest in studio, associate professor at Witshed of what we call RAIL, as well as one of the founders of the deep learning in Dauber. Benji Rossman, we're going to find out about humans and neural networks and stuff. Can you imagine putting my brain in-- just I digress. But can you imagine putting my neural network into a computer? That thing would fry. I don't even know what I'm doing half the time. What's involved? I'm your host, David Watts. My guest in studio is head of robotics, RAIL, autonomous intelligence, and learning, and all sorts of fancy big words. Benji Rossman, thank you, Benji, once again for being here, taking time out. I know you are generally a very, very busy person. Because I mean, you're also a member of the faculty, as we mentioned, of singularity, you, et cetera, et cetera. So here we go. We mentioned before the break, OK? Somebody wants to know, putting a human neural network in a computer, is it such a good idea?

To me, I immediately went, ooh, immortality. Then I'm like, do you really want to put this consciousness in a computer and then give it access to the internet? Probably not. What is your thoughts on that? So actually, there's some interesting questions that go back quite a long way there. So for a bit of context on answering this question, there were big breakthroughs that happened in the area of machine learning in what's now called deep learning, or the older name for it is artificial neural networks. And this really changed everything from the early 2000s. It was built on some ideas that had been around for a while. It was based on our 1940s understanding of how the human brain works. And these artificial neural networks are now the one technique in machine learning that powers most of the big breakthroughs that we're seeing now. And so this is based on basic ideas of how the brain works, that there's all these neurons or these brain cells that pass signals between each other. And it turns out with this very basic idea and some different ways of training these systems, and what we mean by training or when they learn, we're changing the strengths of the connections between them. And this really has powered all of, say, our modern translation systems and image recognition systems and so on.

And as I say, this is very much based on an understanding of the brain. So the trick is to really figure out what the right structure for these things is and how they should learn correctly and so on. So we're inspired very much by natural brains. As far as uploading a human brain or consciousness goes, I think we're still a while away from that. We're still trying to figure out how to get the machine ones right before we interface them. But there's a lot of people doing work on this at the moment, too. But I mean, I think when you talk about that, what I think is amazing in terms of what we're able to get machines to do is for people who have physical problems. And as far as sight is concerned, things like that. Because I know there's some amazing work happening with that at the moment. Yeah, I think it's a bit of a joke. But no doubt, somebody is going to want to be immortalized on a computer.

But another message just in, somebody is wanting you to write-- and they say write an AI. I'm assuming it's a program that customizes adverts to specific people so they don't hear inappropriate ads when their daughters are in the room with them. Hey, Siri, can you help? Somebody else said, this is interesting. Maybe you can give me an answer to this one. Chris reckons-- I can't imagine there's a lot of funding for machine learning in South Africa. How is the rail lab funded? How many students do you have? And are there ways to assist the lab with funding? I'm sure there are ways to assist the lab with funding. Yeah, sure. We're happy to talk to anybody. Happy to talk about funding. Absolutely. Because I mean, Chris must be right. I'm sure. Is there a lot of money in this at the moment? So this is a challenge that typically the kinds of things that we look for funding to fund postgraduate students. So the bulk of the people working in the lab are our postgraduate students. There's that. There's equipment, which is mainly either compute things like GPUs or any powerful computers that can actually run the kinds of algorithms we're developing. And actually, a big thing for us is conference travel.

Because as I mentioned earlier, I'm very passionate about making sure that Africa is a part of the global machine learning community. And so it's really important that our students interact with researchers globally. We've actually got quite a few students in our lab. And actually, we're within the Department of Computer Science and Applied Maths at WIT. And there's even more students there. But I reckon we've probably got about 40 to 50 masters and PhD students that we're working with at the moment. So it's a really big group of really passionate people working on a wide variety of projects, which is really exciting. And it's really been a dream of mine to get to the point, ultimately, that someone somewhere in the world can say, I want to work on machine learning or AI. And someone else will say, have you heard about Johannesburg in South Africa? That's where you should do these things. That's my dream. Because I mean, let's talk about this. Funding is a problem. I mean, labs and things like you guys do does rely quite heavily on funding for it to happen. And if people can get involved with that, I think it would be an absolutely brilliant thing to do. Because you blew me away now when you said you got about 40 or 50 people that work there.

Because to so many of us, this whole deep learning, AI, robotics thing, it is such a foreign concept. You can't imagine what people are doing. We're creating-- I mean, I got invited. I was very, very lucky that I got invited to the launch of the Fox Network's new "War of the World" series, which I think it's just started. It's going to be on DS TV. Got invited there last week. But that's what you think people spend their time doing. Sort of planning on building these robots and weapons of mass destruction and everything. Because I don't know. Is it a human thing? We all think of the bad stuff first and not the good stuff. Because I would imagine there is good stuff that you guys are doing that you actually take for granted, that you haven't even sort of got-- oh, yeah. Yeah. Which we would find fascinating. I think that I do think most of what I do is fascinating. You do. But we don't understand it. I mean, when I look at you and I talk to you in the passion that you talk about these things, it's absolutely amazing. And this deep learning in DARBA, that you've now-- you've been one-- you are one of the founders of it. It's now gone across Africa. How many countries did you say you guys are now sort of got involved in Africa?

So we have this as a-- so we started in 2017, and we've been running this as an annual week-long summer school for people that are already in the field, and usually master's level and above. So we've been trying to get people from all across Africa. I think this year-- so we had the third event a few weeks ago, which was just outside Nairobi. It was the first big one outside of South Africa. And I think we had people from about 40 African countries there. And in fact, we've been running a series of satellite events. Which we call the "Andaba X" series, kind of akin to the TEDx books. And there we've had events in 27 different African countries, which is half of Africa. So these are locally run and organized machine learning training events for people that are in the field to some extent or another. And so it's really exciting. We've got half the continent that's coming together to form a network of practitioners and experts. And it's amazing, because I haven't seen anything like this before. And I don't know how many other continents have 27 countries running machine learning events. Yeah, that's just fascinating. Where do we stand in terms of the rest of the world? I mean, because we're always seen as, oh, Africa, shame, poor Africa, poor starving Africa, poor backwards dock continent Africa.

Where do we stand in terms of this machine learning and deep learning? Do we stack up against the international guys? It's hard to quantify this. And there's a number of ways you could do that. But when we started this in 2017, one of the ideas behind it was that if you take a look at the big machine learning academic conferences around the world, there's a big one in particular that we couldn't find any evidence of any African institution having ever published a paper there. And so we started this. And now subsequently last year, my lab and another lab were the first two from Africa to have papers there. And there have been more papers this year. And we've been sending delegations every year. So now I feel like we're becoming a part of the global community. And that's really exciting because I think there's a lot of great contributions from here that we're now pushing out and other people can get involved in. And we're really collaborating at that level now. It's much smaller than we'd like it to be. But it definitely steps in the right direction. So what is the most exciting thing that you are working on now for you personally? For me personally. Well, there's a couple technical projects.

And there's a couple application projects because I can't think of which one I like the most off the top of my head. Doesn't matter. Share them all. Cool. So in terms of technical ones, one of them that I think maybe a bit easier to describe is working with human robot collaboration. So we've been talking about Siri a bit and this idea that you can yell at Siri and demand whatever it is you need. But what if your robot or your phone or whatever machine it was could anticipate what you wanted by watching you or learning from your behavior? So maybe if you had a little robot in your kitchen and saw you climbing out of bed and shuffling into the kitchen, it could figure out like, oh, from what you normally do on a Saturday morning, you're probably going to go make breakfast. Let me go and turn the stove on or pour some orange juice or something like that to try and understand what humans might want before they ask for it. You could also imagine this kind of in a workshop setting where someone's trying to do something complicated and finicky. And if you had an assistant there that would say, oh, it looks like you're going to need this kind of spanner soon. That could save time and make life easier.

Man, you just took me straight back to my childhood with my dad when he was working on something. I want a number 16 spanner or a number 10 spanner. I'm going, I don't know where it is. I don't know what you want. And then I jumped straight from there to like, man, a robot butler. Jeeves. I mean, that's like the geek dream, right? Oh, you have no idea. Hey, dishes done. Floor done. What would you like for dinner tonight, David? Well, Jeeves, I think I'd like-- sorry, the fridge is out of fresh cream at the moment. Is that a pipe dream? To some extent. But actually, that inspires a lot of the kinds of technical questions we're asking. So an area that we work on a lot in our lab is what we call transfer learning. And this is this idea that if you look at most AI systems that are around these days, they're very good at doing one specific thing. And I've always been more interested in this idea of a more general intelligence or a broader AI, as we call it. So imagine you had your pipe dream butler Jeeves robot, and you taught it how to mop the floor. And now you wanted to learn to sweep the floor. So maybe you demonstrate that or teach it some way. But it should be easier to learn to sweep the floor because you've already taught it to mop the floor than if it hadn't already learned about mopping the floor.

Because there's things like not bumping into walls. There's how you hold whatever tool you're using. There's making contact with the floor and so on. And then if you've then learned to sweep the floor, you should now be better at mopping because you've got more different experience. So this kind of question is something we do a lot of work on. And a lot of our theoretical thinking in the space is around versions of this question. OK.

I'm just looking at some of the questions coming in here. It's absolutely amazing the kind of stuff that can be done. In terms of a lot of people are asking me about machine learning, et cetera, et cetera. People seem very concerned that this is going to take over our lives and take away so much from us. And is there any way-- I mean, you must get people that ask you the same questions. How would you put people's mind at ease? This is a difficult question. And actually, something that I think a lot of us in the community are grappling with, because we've seen a lot of ways that technology has been abused. And personally, my feeling is there's a lot of questions around this that I am not the right person to answer these. And I feel an important part of my job is to get more involvement from people in different sectors and in the public. There's certain questions that we need to answer. I mean, even think about, say, human values. Now, how do we build systems that have the right kind of values? And what do we do when these values differ from country to country? And there have been quite a lot of experiments now that have shown these differences. So how do you build your system-- not even how, but what kind of value should be in the system?

And again, I don't think it should be us working on the technologies that answer these questions. It should be we need more input from people that work in philosophy and in ethics and in sociology and anthropology and all sorts of different areas to actually come together and figure out what do we want to preserve and what do we want to enforce as values and promotes and so on. And what are things we should try and work out of the system? Because I've recently been watching a series-- I think it was on Netflix-- called Unnatural Selection, which is all about gene editing and gene drives, et cetera, et cetera. The stuff that the researchers and the scientists say this stuff can do is absolutely mind-blowing. And then it's like, OK, but then there's the other side of the coin. But I suppose for me to make it more understandable, it goes back to something like a knife or a firearm kind of question. In and of itself, it is not a bad thing. It is the use that you put it to. You can use a knife as a scalpel, as a surgeon. There's many different things. But we don't want to automatically ban all knives. It might be oversimplified. But don't point out the guys who are working on the technology.

Because you're right. I mean, the morals and the ethics of it is you're saying this is technology that we are developing. And it may be sad to say, but if you don't do it-- and thank goodness somebody like you is doing it-- that has these kind of morals and ethics. And you think about the bigger picture. Somebody else is going to. So we need to make a plan there. We're starting to run out of time. Benji, two last questions. What do you see happening in terms of technology in general in South Africa? Going forward, let's talk in a couple of years' time. It's interesting. I've seen a lot of uptake or attempts at uptake from various corporates. And I think that's interesting. I think there's a lot of ways that there could be better services provided by kind of streamlining the ways things work. Not even necessarily with things like artificial intelligence, but just better use of apps and online technology and so on. I'm a big believer that these kinds of systems give people a lot of freedom. And we've obviously seen this in things like cell phones that have enabled complete revolutions in the way we communicate. In the same way, I'm very excited about how we might be able to improve things like health care and education.

In a situation where there's a lot of-- we've got huge shortfalls in people that can provide medical services in a lot of places around the country. And if we can assist the people there by having smarter apps and tools and so on that can help do emergency diagnoses and help connect people with the right kinds of experts, I think we can start making a dent on some of the big problems, like access to these kinds of services around the country. So that's something I'm very excited about. That for me is incredibly exciting, is that-- and then you mentioned-- I think we were chatting off here. You mentioned in terms of education, if you could develop something where the app or the tool or whatever learned and was able to see and assess how you as the learner or the student were learning and where your shortcomings were. Because right now, I mean, a lot of people-- I don't think it's a trend that's going to go away-- are turning towards online to learn. And they're having these short content courses, lifelong learnings becoming a thing. People my age, younger, are going, you know, maybe the job's not so secure. We need to be able to upskill ourselves. But a lot of the stuff, it's presented to you in one format.

And if you're not good at learning that way, you're going to struggle. Now, you said there's something that you guys would be looking at. So almost imagine you could watch a few minutes of an online video, and then it asks you some questions. And if you got them right, it showed you the rest of it. And maybe if you got it wrong, it could either show you the same content taught by someone else in a different way. Or after a few attempts at that, understand, maybe there's some underlying concept you're not getting. And we can go back in the syllabus to that point and recap some of that. And maybe over time, you could learn, oh, this person seems to learn better when there's more visual examples, or when it's more grounded in finance, or when the applications have to do with music, right? So you should be able to learn to customize the way you deliver content to an individual that's best suited to the way that individual learns and interfaces with that content. So I've got this other dream where I'd love to be able to have a system that you could take any person and any content, and it will deliver the curriculum that will give you the highest probability of success of that person learning that content.

See, to me, that is massive. I mean, the day that something like that gets fair-- because I know I'm predominantly-- surprise, surprise, I'm on radio-- an auditory kind of person. So you show me a bunch of graphs, I switch off. That's me done. I do not get fascinated by spreadsheets at all. Probably why I never went into your field, because the numbers and formulas and stuff like that. Benji, we're out of time. In closing, if somebody wants to get hold of you, they want to chat a little bit more, they want to find out about what you guys are doing at VITS, or maybe with regards to some funding. What's the best way to get hold of you? Well, you could visit our website, which is raillab.org. You could tweet at me, which is @BenjaminRosman, or just reach out through VITS University. OK, so it's raillab.org. So there we go, R-A-I-L-L-A-B, raillab.org. Drop them a mail there, Benji, or get hold of it. And I'm sure you'd love to share some of his knowledge and experience, maybe a bit about the lab. Maybe you've got somebody who's interested in getting into this. Chat to the guys there, because I promise you, it'll really get you switched on. Benji, thank you so much. In conclusion, what's next for Benji Rosman?

So now we're still trying to grow our networks across Africa. We're trying to plug in more to some of what's happening internationally, trying to coordinate a lot of our efforts. And we're growing the research we're doing. We've been trying to get hold of some fancy new robots, so we can actually take some of these ideas from more theoretical settings to physical settings, and then pushing a lot of these big applications, like the education and health care. Fantastic stuff. Benji, thank you so much for coming over in the chat to me and helping me to understand a little bit more of this brave new world as we make our way into it. That was Benji Rosman, and as I said, RailLab.org. Next up, another special treat. We go from the robotics to the human side of things. I'm going to be chatting to Pepe Moret about, amongst other things, his life, his work, advertising, and his book called Growing Greatness. It is going to be a goodie. We'll be back with Pepe in just a bit.